Tunnel hole wall detection method and device

Abstract

The invention provides a tunnel cave wall detection method and device, and relates to the field of cave wall detection, and the method comprises the following steps: obtaining and preprocessing multi-modal data in a tunnel, the multi-modal data comprising historical cave wall picture data, a real-time cave wall target image, and environment monitoring data; constructing a cave wall defect detection model, setting an optimization strategy, and training the cave wall defect detection model by utilizing historical cave wall picture data to obtain an optimal cave wall defect detection model; and inputting the obtained real-time cave wall target image into the optimal cave wall defect detection model to obtain a tunnel cave wall detection result. According to the tunnel hole wall detection method and device, the key problems of crack propagation, water seepage, fire disasters and the like in tunnel hole wall detection can be efficiently solved, and technical guarantee is provided for safe operation of a tunnel.

Description

Technical Field

[0001] The present invention relates to the field of tunnel wall detection, and in particular to a tunnel wall detection method and device. Background Art

[0002] Patent CN119414864A discloses an underwater intelligent inspection robot for a water conveyance tunnel and its intelligent inspection method, including the following steps: enabling a robot platform to perform at least one of the maneuvers of advancing, retreating, ascending, and descending in a target water conveyance tunnel according to a target inspection task; monitoring the navigation trajectory and positioning information of the robot platform; receiving real-time instructions from a ground control center, and adjusting the traveling direction and running speed of the robot platform according to the real-time instructions, the navigation trajectory, and the positioning information; collecting muon signals inside the target water conveyance tunnel to generate a surrounding rock density distribution image and leakage channel data according to the muon signals; monitoring obstacles and terrain inside the target water conveyance tunnel to generate a three-dimensional stereoscopic point cloud image of the water conveyance tunnel; performing laser scanning on the tunnel wall of the target water conveyance tunnel to generate a three-dimensional high-resolution point cloud image of the tunnel wall; performing feature fusion processing on the surrounding rock density distribution image, the leakage channel data, the three-dimensional stereoscopic point cloud image of the water conveyance tunnel, and the three-dimensional high-resolution point cloud image of the tunnel wall to obtain a comprehensive tunnel state image and a tunnel data report, and sending the comprehensive tunnel state image and the tunnel data report to the ground control center.

[0003] However, this technology cannot achieve the effect of real-time monitoring and detect the situation of the tunnel wall in real time. Summary of the Invention

[0004] The purpose of the present invention is to provide a tunnel wall detection method and device, which can efficiently address key problems such as crack propagation, water seepage, and fire in tunnel wall detection by real-time monitoring the situation inside the tunnel wall, and provide technical support for the safe operation of the tunnel.

[0005] To achieve the above purpose, the present invention provides a tunnel wall detection method and device, including the following steps: Obtain and preprocess multi-modal data in the tunnel, where the multi-modal data includes historical tunnel wall picture data, real-time tunnel wall target images, and environmental monitoring data; Construct a tunnel wall defect detection model, set an optimization strategy, and train the tunnel wall defect detection model using the historical tunnel wall picture data to obtain an optimal tunnel wall defect detection model; Input the obtained real-time tunnel wall target image into the optimal tunnel wall defect detection model to obtain a tunnel wall detection result.

[0006] Preferably, obtain the preprocessed historical tunnel wall picture data and the real-time tunnel wall target image. Collect the tunnel wall picture data by installing a detection device on the inspection robot, including: Collect the original tunnel wall image and the target image, process the original image and the target image to obtain the RGB values and corresponding pixel coordinates of the original image and the target image.

[0007] Preferably, set an optimization strategy and use the historical tunnel wall picture data to train the tunnel wall defect detection model, including: Compare the result output by the model each time with the real situation. When it is contrary to the real situation, optimize the model parameters and enter a new round of training; When it is the same as the real situation, save the model parameters this time and enter a new round of model training until the correct rate of the model output reaches 98%.

[0008] Preferably, input the obtained real-time tunnel wall target image into the optimal tunnel wall defect detection model to obtain the tunnel wall detection result, including the following steps: Extract the features of the original image and the target image, and then, according to the RGB values of the original image and the target image, perform fusion processing, comparison processing and extraction processing on the original image and the target image, and analyze the defect situation according to the processing results.

[0009] Preferably, extract the features of the original image and the target image, and then, according to the RGB values of the original image and the target image, perform fusion processing, comparison processing and extraction processing on the original image and the target image, including: Perform fusion processing and comparison processing on the original image and the target image to obtain the standard image and shape image of the target image and the target image; Select a pixel coordinate in the original image as the origin, establish a coordinate system, and determine the first shape image coordinate; select the same pixel coordinate as the original image as the origin, establish a coordinate system, and determine the second shape image coordinate. Select multiple points on the shape image and compare the first shape image coordinate and the second shape image coordinate: When the abscissa of the first shape image coordinate and the second shape image coordinate changes, the crack in the tunnel wall expands; When the difference between the ordinates of the first shape image coordinate and the second shape image coordinate is negative, the crack in the tunnel wall expands; When the difference between the ordinates of the first shape image coordinate and the second shape image coordinate is positive, local structural instability may occur in the tunnel wall.

[0010] Preferably, analyzing the defect situation according to the processing results further includes: Add a time series label and a position label to each original image and target image; Each time, multiple pixel coordinates in the original image and the target image are randomly selected, and each pixel coordinate corresponds to an RGB value; According to the timing tag and the position tag, the multiple pixel coordinates in the original image and the target image are compared. Among them, for the pixel coordinates in the original image, the corresponding RGB value is the first RGB value, and for the pixel coordinates in the target image, the corresponding RGB value is the second RGB value. The defect situation is judged by judging the difference between the second RGB value and the first RGB value and the environmental monitoring data.

[0011] Preferably, judging the defect situation by judging the difference between the second RGB value and the first RGB value and the environmental monitoring data includes When the number of cases where the difference between the second RGB value and the first RGB value is greater than or equal to J exceeds P, and the humidity reflected in the environmental monitoring data is greater than a, water leakage occurs on the tunnel wall at this time; When the number of cases where the difference between the second RGB value and the first RGB value is less than J exceeds Q, and there is a combustion factor reflected in the environmental monitoring data, a fire occurs on the tunnel wall at this time.

[0012] A tunnel wall detection device includes: A guide rail, on which an inspection robot is arranged. A data collector is arranged on the inspection robot, and the data collector is used to collect a complete picture inside the tunnel wall; An environmental monitoring sensor is also arranged on the tunnel wall for obtaining environmental monitoring data; A data processing module for preprocessing historical tunnel wall picture data and real-time tunnel wall target images; A model training module for constructing a tunnel wall defect detection model, setting an optimization strategy, and training the tunnel wall defect detection model using the historical tunnel wall picture data to obtain an optimal tunnel wall defect detection model; A detection module for inputting the obtained real-time tunnel wall target image into the optimal tunnel wall defect detection model to obtain a tunnel wall detection result.

[0013] A storage medium stores computer-executable instructions. When the computer-executable instructions are loaded and executed by a processor, the steps of a tunnel wall detection method are implemented.

[0014] Therefore, the present invention adopts the above-mentioned tunnel wall detection method and device, and the technical effects are as follows: Image enhancement and feature extraction: Improve the detectability of defects; Deep learning model: Realize automatic and high-precision defect classification; Multi-modal fusion: Enhance the robustness to complex scenes; Color space and threshold segmentation: Accurately locate pixel-level defects; Geometric Transformations and Coordinate Systems: Ensuring the Spatial Accuracy of Detection Results. Brief Description of the Drawings

[0015] Figure 1 It is a flowchart of a tunnel wall detection method. Detailed Implementation Manner

[0016] The technical solutions of the present invention will be further described below with reference to the drawings and embodiments.

[0017] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the ordinary meanings understood by those of ordinary skill in the field to which the present invention belongs.

[0018] Embodiment 1 As Figure 1 shown, a tunnel wall detection method includes the following steps: Obtain and preprocess multi-modal data in the tunnel, where the multi-modal data includes historical tunnel wall picture data, real-time tunnel wall target images, and environmental monitoring data; Obtain and preprocess historical tunnel wall picture data and real-time tunnel wall target images. Collect tunnel wall picture data by installing a detection device on an inspection robot, including: Collect the original image and target image of the tunnel wall, process the original image and target image to obtain the RGB values and corresponding pixel coordinates of the original image and target image.

[0019] Build a tunnel wall defect detection model, set an optimization strategy, and train the tunnel wall defect detection model using historical tunnel wall picture data to obtain the optimal tunnel wall defect detection model; Set an optimization strategy and train the tunnel wall defect detection model using historical tunnel wall picture data, including: Compare the result output by the model each time with the actual situation. When it is opposite to the actual situation, optimize the model parameters and enter a new round of training; When it is the same as the actual situation, save the model parameters of this time and enter a new round of model training until the correct rate of the model output reaches 98%.

[0020] Input the obtained real-time tunnel wall target image into the optimal tunnel wall defect detection model to obtain the tunnel wall detection result, including the following steps: Extract the features of the original image and target image, and then, according to the RGB values of the original image and target image, perform fusion processing, comparison processing, and extraction processing on the original image and target image, and analyze the defect situation according to the processing results.

[0021] Feature extraction is performed on the original image and the target image. Then, based on the RGB values of the original image and the target image, fusion processing, comparison processing, and extraction processing are carried out on the original image and the target image, including: Fusion processing and comparison processing are carried out on the original image and the target image to obtain the target image, the standard image of the target image, and the shape map; Select a pixel coordinate in the original image as the origin, establish a coordinate system, and determine the first shape map coordinate; select the same pixel coordinate as the original image as the origin, establish a coordinate system, and determine the second shape map coordinate. Select multiple points on the shape map and compare the first shape map coordinate and the second shape map coordinate: When the abscissa of the first shape map coordinate and the second shape map coordinate changes, the crack in the hole wall expands; When the difference between the ordinates of the first shape map coordinate and the second shape map coordinate is negative, the crack in the hole wall expands; When the difference between the ordinates of the first shape map coordinate and the second shape map coordinate is positive, local structural instability may occur in the hole wall.

[0022] Analyze the defect situation based on the processing results, and also include: Add time series tags and position tags to each original image and target image; Each time, randomly select multiple pixel coordinates in the original image and the target image, and each pixel coordinate corresponds to an RGB value; Compare the multiple pixel coordinates in the original image and the target image according to the time series tags and position tags. Among them, for the pixel coordinate of the original image, the corresponding RGB value is the first RGB value, and for the pixel coordinate of the target image, the corresponding RGB value is the second RGB value. Judge the defect situation by judging the difference between the second RGB value and the first RGB value and the environmental monitoring data.

[0023] Judge the defect situation by judging the difference between the second RGB value and the first RGB value and the environmental monitoring data, including When the number of cases where the difference between the second RGB value and the first RGB value is greater than or equal to J exceeds P, and the humidity reflected in the environmental monitoring data is greater than a, water leakage occurs in the hole wall at this time; When the number of cases where the difference between the second RGB value and the first RGB value is less than J exceeds Q, and there is a combustion factor reflected in the environmental monitoring data, a fire occurs in the hole wall at this time.

[0024] A tunnel hole wall detection device includes: A guide rail, on which an inspection robot is arranged, and a data collector is arranged on the inspection robot. The data collector is used to collect complete pictures inside the hole wall; An environmental monitoring sensor is also arranged on the hole wall for obtaining environmental monitoring data; A data processing module for preprocessing historical tunnel wall image data and real-time tunnel wall target images; A model training module for constructing a tunnel wall defect detection model, setting an optimization strategy, and training the tunnel wall defect detection model using the historical tunnel wall image data to obtain an optimal tunnel wall defect detection model; A detection module for inputting the acquired real-time tunnel wall target image into the optimal tunnel wall defect detection model to obtain a tunnel wall detection result.

[0025] Since the embodiments of the device part correspond to the embodiments of the method part, please refer to the description of the embodiments of the method part for the embodiments of the device part, which will not be elaborated here.

[0026] This application also provides a storage medium with a computer program stored thereon, and when the computer program is executed, the steps provided in the above embodiments can be implemented. The storage medium may include: various media such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs that can store program codes.

[0027] Therefore, by adopting the above-mentioned tunnel wall detection method and device, the present invention can effectively address key issues such as crack propagation, water seepage, and fire in tunnel wall detection, providing technical support for the safe operation of tunnels.

[0028] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and these modifications or equivalent replacements do not enable the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A tunnel wall detection method, characterized in that, It includes the following steps: Obtain and preprocess the multi-modal data in the tunnel, where the multi-modal data includes historical tunnel wall picture data, real-time tunnel wall target images, and environmental monitoring data; Construct a tunnel wall defect detection model, set an optimization strategy, and use the historical tunnel wall picture data to train the tunnel wall defect detection model to obtain the optimal tunnel wall defect detection model; Input the obtained real-time tunnel wall target image into the optimal tunnel wall defect detection model to obtain the tunnel wall detection result.

2. The tunnel wall detection method according to claim 1, characterized in that, Obtain and preprocess the historical tunnel wall picture data and real-time tunnel wall target images. Collect the tunnel wall picture data by installing a detection device on the inspection robot, including Collect the original tunnel wall image and the target image, and process the original image and the target image to obtain the RGB values and corresponding pixel coordinates of the original image and the target image.

3. The tunnel wall detection method according to claim 1, characterized in that Set an optimization strategy and use the historical tunnel wall picture data to train the tunnel wall defect detection model, including: Compare the result output by the model each time with the real situation. When it is contrary to the real situation, optimize the model parameters and enter a new round of training; When it is the same as the real situation, save the model parameters of this time and enter a new round of model training until the accuracy rate of the model output reaches 98%.

4. The tunnel wall detection method according to claim 1, wherein Input the obtained real-time tunnel wall target image into the optimal tunnel wall defect detection model to obtain the tunnel wall detection result, including the following steps: Extract the features of the original image and the target image, and then, according to the RGB values of the original image and the target image, perform fusion processing, comparison processing, and extraction processing on the original image and the target image, and analyze the defect situation according to the processing results.

5. A tunnel wall detection method according to claim 1, characterized in that Extract the features of the original image and the target image, and then, according to the RGB values of the original image and the target image, perform fusion processing, comparison processing, and extraction processing on the original image and the target image, including: Perform fusion processing and comparison processing on the original image and the target image to obtain the standard image and shape map of the target image and the target image; Select a pixel coordinate in the original image as the origin, establish a coordinate system, and determine the first shape map coordinate; select the same pixel coordinate as the original image as the origin, establish a coordinate system, and determine the second shape map coordinate. Select multiple points on the shape map and compare the first shape map coordinate and the second shape map coordinate: When the abscissa of the first shape map coordinate and the second shape map coordinate changes, the crack in the tunnel wall expands; When the difference in the ordinate between the first shape map coordinate and the second shape map coordinate is negative, the crack in the tunnel wall expands; When the difference in the ordinate between the first shape map coordinate and the second shape map coordinate is positive, local structural instability may occur in the tunnel wall.

6. The tunnel wall detection method according to claim 5, characterized in that, Analyze the defect situation according to the processing results, and also include: Add a time series label and a position label to each original image and target image; Randomly select multiple pixel coordinates in the original image and the target image each time, and each pixel coordinate corresponds to an RGB value; Compare multiple pixel coordinates in the original image and the target image according to the timing tag and the position tag. Among them, for the pixel coordinates of the original image, the corresponding RGB value is the first RGB value, and for the pixel coordinates of the target image, the corresponding RGB value is the second RGB value. Determine the defect situation by judging the difference between the second RGB value and the first RGB value and the environmental monitoring data.

7. A tunnel wall detection method according to claim 6, characterized in that, Determine the defect situation by judging the difference between the second RGB value and the first RGB value and the environmental monitoring data, including: When the number of differences between the second RGB value and the first RGB value greater than or equal to J exceeds P, and the humidity reflected in the environmental monitoring data is greater than a, water leakage occurs on the tunnel wall at this time; When the number of differences between the second RGB value and the first RGB value less than J exceeds Q, and there is a combustion factor reflected in the environmental monitoring data, a fire occurs on the tunnel wall at this time.

8. A tunnel wall detection device, characterized in that Including: A guide rail, on which an inspection robot is arranged. A data collector is arranged on the inspection robot, and the data collector is used to collect complete pictures inside the tunnel wall; An environmental monitoring sensor is also arranged on the tunnel wall for obtaining environmental monitoring data; A data processing module for preprocessing historical tunnel wall picture data and real-time tunnel wall target images; A model training module for constructing a tunnel wall defect detection model, setting an optimization strategy, and training the tunnel wall defect detection model with historical tunnel wall picture data to obtain an optimal tunnel wall defect detection model; A detection module for inputting the obtained real-time tunnel wall target image into the optimal tunnel wall defect detection model to obtain a tunnel wall detection result.

9. A storage medium, characterized in that, Computer-executable instructions are stored in the storage medium. When the computer-executable instructions are loaded and executed by a processor, the steps of a tunnel wall detection method according to any one of claims 1 to 7 are implemented.

Images